Major Histocompatibility Complex Class I-related Chain A and UL16-Binding Protein Expression on Tumor Cell Lines of Different Histotypes

INTRODUCTION

NK 3 cells play a relevant role in host defense, because they represent the most efficient mechanism used by the immune system to rapidly detect and destroy tumor and virally infected cells (1 , 2) . NK cells spare normal cells expressing adequate amounts of MHC class I molecules, whereas they kill transformed cells that have down-regulated or even lost the expression of MHC class I molecules (3) . The discovery of MHC-specific inhibitory receptors clarified the molecular basis of this major NK cell function. These inhibitory receptors belong either to the immunoglobulin superfamily (KIR) or to the C-type lectin superfamily (CLRs; Refs. 4, 5, 6, 7 ).

More recently, the receptors involved in triggering of NK cell-mediated cytotoxicity have been unveiled (8, 9, 10) . These are represented by three novel NK-specific receptors, termed NKp46, NKp44, and NKp30 (11, 12, 13, 14, 15) , that belong to the immunoglobulin superfamily and represent the first members of a group of receptors collectively termed NCRs (16) . NCRs associate with different immunoreceptor tyrosine-based activating motif-bearing signal transducing polypeptides, including CD3ζ, FcεRIγ, and KARAP/DAP12 (12 , 14 , 15) , that allow the delivery of intracellular signals resulting in induction of cytolytic activity. NKp46 and NKp30 are expressed by both resting and activated NK cells (11 , 15) , whereas NKp44 is expressed only on activated NK cells (13) . Phenotypic analysis of the NCR surface density has showed that it can vary in different individuals, and that there is a coordinated expression of the three receptor molecules at the clonal level, leading to a NCR^dull or NCR^bright phenotype (15) . A direct correlation exists between the surface density of NCR and the ability of NK cells to kill certain tumors (17) . Moreover, NK cell-mediated cytotoxicity can be virtually abrogated by the simultaneous masking of the three NCR (15 , 17) . The nature of the cellular ligands recognized by NCR on tumor or normal cells is still undefined, although putative viral ligands have been reported (18 , 19) .

Another receptor involved in NK cell-mediated cytotoxicity is represented by NKG2D, a C-type lectin molecule (20, 21, 22) . This receptor was shown to associate with the signaling molecule DAP10/KAP10 (21 , 23) , which recruits phosphatidylinositol 3-kinase (21 , 23 , 24) . Unlike NCR, the expression of NKG2D is not confined to NK cells because this molecule is also detected on virtually all human TCR γ/δ⁺ and CD8⁺ TCR α/β⁺ cells. Moreover, NKG2D expression appears relatively homogeneous in both NCR^bright and NCR^dull NK cells (22) . Recent data indicate that in most instances NK cell triggering in the process of tumor cell lysis depends on the concerted action of NCR and NKG2D. Thus, the combined masking of NCR and NKG2D was found to abrogate the NK-mediated lysis of most tumor target cells. On the other hand, in NCR^dull cells NKG2D played a major role as a triggering receptor. Thus, NCR^dull NK clones (or NK cells pretreated with anti-NCR mAbs) can be used as suitable effector cells to evaluate the NKG2D-dependent lysis (22) .

The target cell ligands for NKG2D are represented by MICA and MICB molecules (20 , 21 , 25) . These are stress-inducible molecules encoded within the human MHC complex that are mostly expressed on tumors of epithelial origin (26, 27, 28) , but also certain melanomas have been described to express MICA (22 , 29) . Moreover, functional data using MICA/B negative targets (such as the Daudi Burkitt’s lymphoma) suggested that additional NKG2D ligands may exist (22) . Treatment of these target cells with appropriate enzymes suggested that such additional NKG2D ligands might be represented by GPI-linked molecules (22) . In this context, a novel family of human, GPI-linked, MHC class I-related molecules, termed ULBPs could bind to NKG2D and activate NK cells (30, 31, 32) . Notably, in mice, the Rae1 molecule, that together with H60 represents the major ligand for NKG2D, is also a GPI-linked protein (33 , 34) . Recent studies have indicated that these ligands for murine NKG2D are induced on some tumor cells and that their ectopic expression mediates potent NK- and T-cell-mediated rejection of MHC class I-bearing tumors in vivo (35) .

Because NK cell function is the result of a fine tuning between opposite signals by inhibitory and/or activating receptors, it is conceivable that the final signal delivered to a given NK cell may depend on the relative number of the various receptor/ligand interactions and on the avidity of each receptor engagement. Indeed, in humans it has been proposed that even HLA class I⁺ target cells may be killed by NK cells via NKG2D (20 , 30) . This event, however, is likely to occur only when the appropriate ligands for this receptor are up-regulated at the cell surface. Given these recent findings, it is tempting to speculate that, similar to Rae1 and H60 in the mouse, MICA/B and ULBPs may stimulate human antitumor responses in vivo.

In the present study, we analyzed a large panel of tumor cell lines of different histotypes for the surface expression of MICA and ULBPs. Moreover, the functional relevance of MICA and ULBP expression by tumor cells was evaluated in functional studies using activated or freshly isolated NK lymphocytes as effector cells.

MATERIALS AND METHODS

Tumor Cell Lines.

The human tumor cell lines analyzed in this study were all grown in RPMI 1640 supplemented with 10% FCS. The melanoma cell lines included the β2μ-deficient cell lines 1074mel, 1106mel, 1174mel, 1259mel, FO-1, Me 1386, Me 18105, MEL39, and SK-MEL-33 (36, 37, 38, 39) . AUMA and IRNE were generous gifts from Dr. P. Coulie (Université Catholique de Luovain, Brussels, Belgium). The HLA-class I⁺ melanoma cell lines included Colo38, DM391, M14, T372A, and MEL15 (40, 41, 42) and the two cell lines derived from metastases resected at different times from patient LB33, MEL.A, and MEL.B (43) . The T-cell lines analyzed were the T-lymphoblastoid CCRF-CEM (referred in the text as CEM), the T-cell lymphoma H9, the T-cell leukemias CCRF-HSB-2 (referred in the text as HSB-2) and Molt-4 (ATCC CCL 119, HTB-176, CCL-120, and CRL 1582, respectively), and the Jurkat cell clone JA3 (22) . The B-CLL MEC1 and MEC2 were derived from a patient on two subsequent occasions (44) . The LCL cell lines 721.221 and C1R, the Burkitt’s lymphoma cell lines Daudi and Raji (ATCC CRL-213 and CCL 86, respectively), and the B lymphoma Akuba and Silfere (kindly provided by Dr. P. Fisch, Institute of Pathology, Freiburg, Germany) were also analyzed. The AML cells included Mono Mac 6 (referred in the text as MM6) cell line (45) and the leukemic cells GT 8, GT 11, and GT ER (classified as M3, M5 and M7, respectively, subtypes according to the French-American-British classification), derived from peripheral blood collected upon informed consent at the San Martino Hospital (Genova, Italy). We also analyzed the LAN5 and SK-N-BE neuroblastoma cell lines (ICLC HTL 96022 and HTL 96015, respectively); the HT-29, WiDr, and SW480 colon carcinoma (ATCC HTB-38, CCL-218 and CCL-228, respectively); the IGROV1 (46) and OC315 (ICLC HTL 98013) ovarian carcinoma; the HeLa cervical carcinoma (ATCC CCL-2); and the SMMC hepatocellular carcinoma (kindly provided by Dr. H. Xie, Shanghai Institute of Cell Biology, Shanghai, China) cell lines.

Purification of NK Lymphocytes and Generation of Polyclonal NK Cell Populations or Clones.

Peripheral blood lymphocytes were derived from healthy donors by centrifugation on a Ficoll-Hypaque gradient and depletion of plastic-adherent cells. To obtain enriched NK cells, peripheral blood lymphocytes were incubated with anti-CD3 (JT3A), anti-CD4 (HP2.6), and anti-HLA-DR (D1.12) mAbs (30 min at 4°C), followed by goat antimouse IgG-coated Dynabeads (Dynal, Oslo, Norway; 30 min at 4°C) and immunomagnetic depletion (13) . CD3⁻, CD4⁻, HLA-DR⁻ cells were cultured on irradiated feeder cells in the presence of 100 units/ml recombinant interleukin 2 (Proleukin; Chiron Italia srl, Milan, Italy) and 1.5 ng/ml phytohemagglutinin (Life Technologies, Inc., Paisley, Scotland) to obtain polyclonal NK cell populations or, after limiting dilution, NK cell clones. When NK cells were tested unstimulated as effector cells in cytotoxicity assays, they were alternatively isolated from peripheral blood using the RosetteSep method (StemCell Technologies, Vancouver, British Columbia, Canada).

mAbs and Flow Cytometric Analysis.

To characterize the surface expression of the NKG2D ligands on the tumor cell lines, the following mAbs were used: BAM195 (anti-MICA, IgG1; IST, Genova, Italy; Ref. 20 ); M295 (anti-ULBP1, IgG1), M310 and M311 (anti-ULBP2, IgG1), M550 and M551 (anti-ULBP3, IgG1), and M362 (anti-MICB, IgG1; Immunex, Seattle, WA; Ref. 30 ). To evaluate surface expression of HLA-class I, W632 mAb was used (42) .

NK cell lymphocytes were phenotypically characterized by the use of BAB281 (anti-NKp46, IgG1; Ref. 11 ), A76 (anti-NKp30, IgG1; Ref. 15 ), Z231 (anti-NKp44, IgG1; Ref. 13 ), BAT221 (anti-NKG2D, IgG1; Ref. 22 ), c127 (IgG1, anti-CD16), and c218 (IgG1, anti-CD56; Ref. 15 ).

Cells were stained with the appropriate mAb, followed by phycoerythrin-conjugated, isotype-specific, goat antimouse second reagent (Southern Biotechnology Associated, Birmingham, AL). Samples were analyzed by cytofluorometric analysis on a FACSort and with the Cell Quest program (both from Becton Dickinson, Mountain View, CA).

Cytolytic Assays.

Cells were tested for cytolytic activity in a 4-h ⁵¹Cr-release assay as described previously (15) , either in the absence or in the presence of various mAbs. For masking experiments, the anti-NCR mAb of IgM isotype were used: the anti-NKp46 KL247, the anti-NKp44 KS38, both kindly provided by S. Parolini (University of Brescia, Brescia, Italy; Ref. 15 ), and the anti-NKp30 F252 mAb (47) . To assess the involvement of MICA and ULBPs on NKG2D-dependent lysis, the effector cells were first incubated with the anti-NCR mAb (for 15 min at room temperature in the cytotoxicity plate) to block these triggering receptors. Then target cells were added followed by one or another anti-ligand mAb: BAM195 (anti-MICA), M295 (anti-ULBP1), M311 (anti-ULBP2), and M551 (anti-ULBP3). Masking of NKG2D with BAT221 mAb was also performed.

The concentration of the various mAbs was 10 μg/ml or otherwise specified. The E:T ratios are indicated in the text. The effector cells were either freshly derived NK cells or polyclonally activated NK cells.

Reverse Transcription-PCR Analysis and MICA Allele Determination.

Total RNA was isolated from melanoma cells using TRIZOL reagent (Invitrogen, Carlsbad, CA), following the manufacturer’s instructions. Two μg of total RNA were then reverse transcribed into cDNA with Moloney murine leukemia virus reverse transcriptase (Invitrogen) and oligo-dT primers (Invitrogen) in a 20-μl reaction. One μl of cDNA was then subjected to PCR using MICA-, MICB-, and GAPDH-specific primers. The sequence of the primers used is: MICA fu 5′, 5′-ATGGGGCTGGGCCCGGTCTTC-3′, MICA tm 3′, 5′-AGCAGAAACATGGAATGTCTGCCAA-3′; MICB fu 5′, 5′-ATGGGGCTGGGCCGGGTCCTGCTGTTT-3′, MICB tm 3′, 5′-AGAAACATATGGAAAGTCTGTCCGT-3′; and GAPDH forward, 5′-TGAAGGTCGGAGTCAACGGATTTGGT-3′, GAPDH reverse, 5′-CATGTGGGCCATGAGGTCCACCAC-3′. The PCR conditions for MICA and MICB amplifications were 95°C for 2 min hot start, 95°C for 30 s, 59°C for 30 s, and 72°C for 1 min, for 30 cycles, and for GAPDH, amplifications were 95°C for 2 min hot start, 95°C for 30 s, 60°C for 30 s, and 72°C for 1 min for 20 cycles. The PCR products were then run on a 1.5% agarose gel and visualized by ethidium bromide staining. For MICA allele determination, DNA bands of the predicted size were cut out of the gel, and DNA fragments were purified using a QIAquick gel extraction kit (Qiagen, Valencia, CA). MICA DNA fragments were then subjected to sequencing, and the sequences were compared with the MICA database provided by the National Center for Biotechnology Information (Bethesda, MD). MICA alleles could also be determined by identifying the nucleotides at the bimorphic positions in the sequence.

RESULTS

Expression of MICA and ULBPs in Melanoma Cell Lines.

The pattern of surface expression of the NKG2D ligands MICA and ULBPs was first evaluated in a panel of 18 melanoma cell lines by the use of specific mAbs. In a previous study, we showed that FO-1 melanoma cells express high surface density of MICA, as assessed by the bright staining with the MICA-specific BAM195 mAb (22) . These data provided unequivocal evidence that MICA expression is not confined to tumors of epithelial origin, as reported previously (26, 27, 28) . On the other hand, MICA was not detected on melanoma cells M14, suggesting that MICA expression can be heterogeneous among different melanoma cell lines. Indeed, as shown in Table 1 ⇓ , the pattern of expression of MICA and ULBP is rather variable in the panel of melanoma cell lines analyzed. For example, FO-1 displayed high levels not only of MICA but also of ULBP2 and ULBP3, whereas other melanomas including M14 expressed low levels of NKG2D ligands. Expression of MICA was detected in the majority of the melanomas analyzed (14 of 18), whereas its surface density displayed major variations. For example, FO-1, 1174mel, and Me 18105 cells were characterized by a high MICA expression, whereas 1074mel and MEL15 were characterized by a rather low expression. We also comparatively analyzed the melanoma cell lines, MEL.A and MEL.B, derived from metastases resected at different times from the same patient and displaying differences in HLA class I phenotype (43) . Indeed, as shown previously, whereas MEL.A expresses a complete set of HLA class I alleles, MEL.B only expresses the HLA-A24 allele. Interestingly, no substantial differences in terms of NKG2D ligands could be detected, and both cell lines displayed the MICA^low ULBP⁻ phenotype (although the level of MICA expression was slightly higher in the first metastasis).

Reverse Transcription-PCR analysis was performed on several melanoma cell lines for both MICA and MICB (Table 2) ⇓ . It can be seen that among the cell lines scored as MICA⁻ by FACS analysis, only Me 1386 was found to be negative by PCR analysis, whereas 1106mel was weakly positive. On the other hand, the Colo 38 and M14 melanoma cell lines were clearly positive by PCR. Sequence analysis revealed that both Colo 38 and M14 express the MICA*010 allele, an allele characterized as carrying an arginine-to-proline substitution at position 6. This substitution renders the protein unstable and highly degradable (48) . Regarding MICB, it is of note that although most melanoma cell lines analyzed were scored as MICB⁺ by PCR, cytofluorometric analysis using the M362 mAb (which reacts with MICB-transfected COS-7 cells) resulted in negative staining in all instances. This suggests that MICB might not be expressed at the cell surface, at least in the melanoma cell panel analyzed.

Expression of MICA and ULBPs in Leukemias.

The cytofluorometric analysis of MICA and ULBP surface expression was also performed in a panel of leukemias. As shown in Table 3 ⇓ , in leukemic T cell lines, MICA was either expressed at low surface density (in JA3 and HSB-2) or was not detectable (in H9, CEM, and Molt-4). On the other hand, ULBP molecules were expressed in all of the cases analyzed (Table 3) ⇓ . In particular, ULBP2 was detected at medium/high levels in all T-cell leukemias, ULBP1 on 3 of 5, whereas ULBP3 was expressed on the H9 cell line only.

Among tumors of the B-cell lineage (Table 3) ⇓ , B-CLL (represented by MEC1 and MEC2) were MICA⁻/ULBP⁻, whereas the Burkitt’s lymphoma cell lines were either MICA⁻ ULBP⁻ (Raji) or ULBP1⁺ only (Daudi). Finally, also LCLs, such as 721.221 and C1R, and the B lymphoma Akuba and Silfere expressed few NKG2D ligands (the only one being represented by ULBP3 on C1R cells). Thus, the B-cell lines are generally characterized by the MICA⁻ phenotype and only in a few cases express a given ULBP molecule. Notably, ULBP2 was detected on all of the T-cell lines tested but not on the B-cell lines tested. Finally, cytofluorometric analysis detected neither MICA nor ULBP molecules on four acute myeloid leukemias (Table 3) ⇓ .

Expression of MICA and ULBPs by Carcinoma and Neuroblastoma Cell Lines.

We also analyzed a panel of carcinoma cell lines (Table 4) ⇓ , including the 3 colon carcinomas HT29, WiDr and SW480, the 2 ovarian carcinomas IGROV1 and OC315, the cervical carcinoma HeLa, and the hepatic carcinoma SMMC. As suggested by previous studies (28) , all of these tumors express MICA; interestingly, they also express ULBP molecules. In particular, ULBP3 is expressed at high levels by the majority of the carcinoma cell lines analyzed. Thus, carcinoma cells express an array of different ligands that are likely to represent a major target for tumor detection by NKG2D-expressing NK and T lymphocytes. Finally, we analyzed the neuroblastoma cell lines SK-N-BE and LAN5 (Table 4) ⇓ . Previous studies indicated that neuroblastoma and glioblastoma cell lines were susceptible to NK-mediated killing and that masking of NCR resulted in a virtual abrogation of cytolysis (49) , thus suggesting a major involvement of NCR but not of NKG2D. In agreement with these functional data, the neuroblastoma cell lines analyzed expressed the MICA⁻/ULBP⁻ phenotype, thus confirming that NKG2D is not involved in the NK-mediated killing of these tumors.

Recognition of Different NKG2D Ligands by Activated NK Cells.

In these experiments, NK cell populations derived from different donors were analyzed for tumor cell lysis either in the absence or in the presence of anti-NKG2D or anti-NCR mAb (i.e., a mixture of anti-NKp30, anti-NKp46, and anti-NKp44 mAbs). As shown in Figs. 1 ⇓ 2 ⇓ 3 ⇓ , blocking of one or another type of receptor resulted in different inhibitory effects, depending upon the source of target cell analyzed. Thus, to directly assess the role of the interaction between NKG2D and its specific ligands in tumor cell lysis, the same effector cells were pretreated with saturating amounts of anti-NCR mAbs to block the interaction of the NCR with their ligands expressed on target cells. Under these experimental conditions, the remaining NK-mediated cytotoxicity was expected to be mainly dependent upon engagement of the NKG2D. Along this line, we showed previously that NK cells expressing low surface densities of NCR (NCR^dull) were still capable of killing various tumor cell lines. Moreover, in this case, cytotoxicity was mainly dependent on the function of NKG2D. Indeed, NKG2D was equally expressed and functional in all types of NK cells, including the NCR^dull ones (22) .

These effector cells pretreated with anti-NCR mAb were assessed for cytotoxicity either in the absence or in the presence of anti-NKG2D, anti-MICA, or anti-ULBP mAb. In Fig. 1 ⇓ , data regarding 3 representative melanoma cell lines are shown. Cytofluorometric analysis (Fig. 1A) ⇓ revealed the following phenotypes: AUMA was MICA⁺ ULBP3^low, MEL39 was MICA⁺ ULBP2^low ULBP3^low, whereas FO-1 was MICA⁺ ULBP2⁺ ULBP3⁺ (see also Table 1 ⇓ ). Accordingly, killing of AUMA or MEL39 cells was virtually abrogated in the presence of either anti-NKG2D or anti-MICA mAb. This indicates that MICA represents the major NKG2D ligand in both melanoma cell lines and that in these tumors the low ULBP surface density does not play a significant role in NKG2D-mediated killing. At variance with these results, in the case of FO-1 cells, the combined use of anti-MICA and a mixture of anti-ULBPs mAbs resulted in a synergistic effect, thus indicating that NKG2D recognized both MICA and ULBP. Taken together, these data confirm the role of NKG2D in the induction of NK-mediated cytotoxicity against melanoma cells and also provide evidence that, in these tumor cells, both MICA and ULBP molecules can trigger NK cells via NKG2D. Consistent with previous data (13 , 17) , killing of melanomas, such as M14, MEL15, and MEL.B that are characterized by a low expression of the NKG2D ligands, was mediated via NCR (not shown).

The results of surface expression in T-cell lines (see Fig. 2A ⇓ and Table 3 ⇓ ) suggested that ULBPs may represent the major NKG2D ligands. Indeed, as shown in Fig. 2B ⇓ , by the use of anti-ULBP mAbs, a strong inhibition of the NK-mediated lysis against the T-lymphoblastoid CEM (MICA⁻ ULBP1⁺ ULBP2⁺) and the T-lymphoma H9 (MICA⁻ ULBP2⁺ ULBP3⁺) cell lines could be detected. On the other hand, the same mAbs were only partially inhibitory when the T-leukemia JA3 cell line was used as a target. In this case, however, target cells also expressed low levels of MICA. Accordingly, the combined use of anti-ULBP and anti-MICA mAb resulted in stronger inhibition of cytotoxicity. The role of NKG2D ligands was then assessed both in Daudi Burkitt’s lymphoma and in C1R LCL, i.e., the only B-cell lines expressing NKG2D ligands.

We reported previously that Daudi Burkitt’s lymphoma was lysed by NK cells in an NKG2D-dependent fashion (22) . The putative ligand was a GPI-linked molecule, as suggested by the finding that the cytolytic activity was abrogated by PI-PLC treatment of target cells. These data suggested a possible involvement of ULBPs in the NK-mediated lysis of this target. Indeed, as shown in Fig. 3 ⇓ , we found that Daudi expressed surface ULBP1 molecules and that anti-ULBP1 mAb could strongly inhibit the cytolytic activity mediated by polyclonal NK cells. In this case the level of inhibition upon ULBP1 mAb-mediated masking was similar to that detected in presence of anti-NKG2D mAb (Fig. 3B) ⇓ . Also, the lysis of C1R cells was mediated via NKG2D upon interaction with ULBP molecules (in this case, ULBP3). However, the NK-mediated killing of these target cells was dependent upon signals generated via both NKG2D and NCR. Thus, inhibition by anti-NKG2D or anti-ULBP mAb was maximal when used in combination with anti-NCR mAb. Fig. 3 ⇓ shows the results of the NK-mediated lysis of another Burkitt’s lymphoma cell line (Raji), which expresses neither MICA nor ULBP molecules. These results indicate that killing of this cell line was essentially NCR dependent. Results comparable with those reported in Fig. 3 ⇓ for the Raji cell line were obtained with a panel of additional B-cell lines including: MEC1, MEC2, 721.221, Akuba, and Silfere (not shown).

In conclusion, the above data indicate that NK-mediated killing of T- or B-cell tumors involves predominantly ULBP1 and ULBP2 as target molecules of NKG2D-mediated recognition. It is of note, however, that although killing of T-cell leukemic cells is predominantly (but not exclusively) mediated via NKG2D, the NK cell-mediated killing of most B-cell lines (6 of 8) is NKG2D independent. Indeed, mAb-mediated masking of NCR resulted in strong inhibition of cytolytic activity against these target cells. Although not shown, the four AMLs analyzed were lysed via NKG2D-independent mechanisms. This is in line with a previous functional report (50) and with their MICA⁻ ULBP⁻ surface phenotype (Table 3) ⇓ .

Correlation of NKG2D Ligand Density on Tumor Cells with Their Susceptibility to Resting NK Cell-mediated Lysis.

We analyzed a panel of HLA class I-negative melanoma cell lines for their susceptibility to lysis mediated by fresh NK cells. We showed previously (15) that, in freshly derived NK cells, NKp30 and NKp46 cooperate in inducing the lysis of certain melanoma cell lines. However, the combined masking of these two NCRs did not result in a complete abrogation of cytotoxicity, thus suggesting the existence of additional receptors involved in target cell lysis mediated by fresh NK cells. As shown above (Table 1) ⇓ , most melanoma cell lines are characterized by high levels of MICA expression. Accordingly, killing of these melanoma cells by fresh NK cells was efficiently inhibited not only by anti-NCR mAb but also by anti-NKG2D mAb in most of the melanomas analyzed (Fig. 4) ⇓ . Moreover, a similar inhibitory effect was obtained by anti-MICA mAb when used in combination with anti-NCR mAbs. In these experiments, an exception was represented by Me 1386 cells that are MICA⁻ ULBP⁺. In this case, anti-MICA mAb had no effect, whereas anti-ULBP mAb strongly inhibited the NKG2D-dependent lysis. Thus, also ULBP alone is sufficient to trigger fresh NK cells via NKG2D. In line with this finding, the killing of various T-cell leukemias by fresh NK cells was mainly dependent upon the interaction of ULBP molecules and NKG2D (not shown).

Remarkably, 2 of 11 HLA class I⁻ melanoma cell lines analyzed were resistant to lysis by fresh NK cells (isolated from different donors). These cell lines (1074mel and 1106mel; Fig. 5A ⇓ ) were characterized by the lack of (or low) surface expression of both MICA and ULBP molecules (see Table 1 ⇓ ). On the contrary, other melanomas such as 1174mel, FO-1 (Fig. 5A) ⇓ , AUMA, and Me 18105 (not shown) were highly susceptible to fresh NK cells and expressed high levels of NKG2D ligands (see Table 1 ⇓ ). Because all of these melanomas were HLA class I negative, the observed differences in susceptibility to lysis cannot be consequent to KIR or CD94/NKG2A interactions with HLA class I molecules. As shown in Fig. 5B ⇓ , activated NK cells also lysed more efficiently 1174mel and FO-1 than 1106mel and 1074mel, although the differences in susceptibility to lysis were less marked.

Other tumor cell lines of different histotypes were also tested for susceptibility to lysis by both fresh and activated NK cells (Fig. 5, B and D ⇓ , respectively). The T-cell lymphoma H9 (ULBP2⁺ ULBP3⁺) and the ovarian carcinoma OC315 (MICA⁺) represented targets characterized by high expression of NKG2D ligands. On the contrary, both the LCL 721.221 and the neuroblastoma SK-N-BE were MICA⁻ ULBP⁻. Killing by resting NK cells was more efficient against H9 and OC315 than against 721.221 and SK-N-BE. This is in line with the above results in which NK cells were tested against melanoma target cells. Thus, at least in fresh NK cells, the interaction between NKG2D and one or another NKG2D ligand appears to correlate with the magnitude of target cell lysis. On the other hand, this correlation is, at least in part, lost when effector cells are represented by activated NK cells. Thus, for example, the lysis of 721.221 target cells by activated NK cells was comparable with that of H9 and OC315, whereas SK-N-BE remained substantially resistant to lysis.

DISCUSSION

In this study, we analyzed the role of NKG2D in human NK cell-mediated lysis of tumor cells expressing different NKG2D-specific ligands. In addition to MICA, we investigated the surface expression and the functional relevance of the recently described ULBP molecules (ULBP1, ULBP2, and ULBP3) on a large panel of tumor cells. Our results indicate that ULBP molecules may or may not be coexpressed with MICA in tumors of different histotypes. Moreover, we show that ULBP molecules are responsible for the induction of NKG2D-dependent killing of a number of MICA⁻ tumors. Importantly, these findings allow us to clarify why, despite the lack of MICA surface expression, certain tumor cells are susceptible to NK-mediated lysis induced via NKG2D. Thus, certain EBV⁺ lymphomas and most T-cell leukemias express little or no MICA. However, they are killed by NK cells via the NKG2D-dependent pathway of activation. This is possible thanks to the expression of NKG2D ligands belonging to the ULBP family. Our study, although confirming the surface expression and the functional relevance of MICA in carcinoma cell lines (26, 27, 28) or in melanomas (22 , 29) , shows that ULBP molecules represent an additional NKG2D ligand in these tumors. In particular, although ULBPs are poorly expressed in melanomas (Table 1) ⇓ , they are frequently expressed on MICA⁺ carcinoma cell lines including lung, colon, ovarian, and cervical carcinomas. Other tumors analyzed such as neuroblastoma, AML, and certain EBV⁺ lines are lacking both MICA and ULBP. Accordingly, these target cells are killed via NCR-dependent mechanisms (17 , 49 , 50) .

Unlike NCRs that are expressed at different surface densities in different NK cells, the expression of NKG2D is relatively homogeneous (22) . This would imply that the activation of a given NK cell via NKG2D may simply be dictated by the expression of appropriate ligands on target cells. Thus, when target cells express low surface densities (or do not express) MICA or ULBP (see Tables 1 ⇓ , 3 ⇓ , 4 ⇓ ), no NKG2D-dependent NK cell triggering occurs. On the other hand, the surface expression of high levels of NKG2D ligands is required for the induction of NK cell triggering, especially in the case of fresh NK cells. Along this line, among HLA class I-negative tumors, the 1174mel and FO-1 melanomas (expressing high levels of NKG2D ligands) were killed efficiently by fresh NK cells, whereas the 1074mel and 1106mel melanomas (expressing little or no ligands) were resistant to lysis (Fig. 5) ⇓ . This correlation between expression of NKG2D ligands and augmented susceptibility to lysis may not always be observed when activated NK cells are used as effector cells. This is likely attributable to the usage of additional triggering receptors, which are expressed only upon NK activation, such as NKp44 (13) . Moreover, this may also depend upon the increased cytolytic potential of NK cells after culture in the presence of recombinant interleukin 2 (10) .

The expression of high levels of MICA (or ULBP) on target cells is likely to also be important to override the negative signaling via KIR molecules (in the case of HLA class I⁺ cells). In this context, triggering via NKG2D may or may not be down-regulated by the engagement of KIR with HLA class I ligands; this reflects the balance between the intensity of negative (KIR) and positive (NKG2D or NCR) signals that occur upon NK cell engagement with target cells. Indeed, it is well known that HLA class I-defective LCLs, such as C1R and 721.221, are killed efficiently by NK cells unless protective alleles are expressed upon cell transfection. Although 721.221 cells are killed via NKp46, the lysis of C1R cells is consequent to the engagement of both NCR and NKG2D. Data on HLA class I cell transfectants characterized by high levels of a given HLA class I allele expression suggest that both NCR- and NKG2D-dependent positive signaling can be down-regulated by inhibitory signals generated upon recognition of HLA class I. On the other hand, high levels of NKG2D ligands on tumor target cells may explain, at least in part, our findings that both HLA class I-positive and -negative tumors were efficiently killed by NK cells via NKG2D.

We also analyzed a panel of melanomas for MICA and MICB mRNA expression and correlated these data to the expression of these molecules at the cell surface. Whereas anti-MICA (BAM 195) reacted with most (but not all) cell lines scored as positive for MICA transcripts, anti-MICB (M362) did not react with any of the melanoma cell lines analyzed. This finding suggests that MICB (although recognized by the specific mAb on MICB transfected COS-7 cells) may not be expressed at the cell surface in tumors scored as MICB⁺ on the basis of PCR analysis. Along this line also, tumors other than melanoma (including the hepatic carcinoma SMMC, the ovarian carcinoma IGROV1, the cervical carcinoma HeLa, the T-cell lymphoma H9, and the Burkitt’s lymphoma Raji) were scored as MICB⁺ by PCR but did not react with anti-MICB mAb (not shown).

Regarding MICA, a lack of correlation between the PCR data and the determination of surface expression was found in Colo38 and M14 cells. Although not shown, functional data indicate that NKG2D is not involved in killing of Colo 38 melanoma (which is also characterized by the ULBP⁻ phenotype). These data are in line with a previous study indicating that the MICA*010 allele is poorly (or is not) expressed at the cell surface (48) . A previous report on the analysis of ULBP transcripts in various cell types suggested a possible lack of correlation between mRNA expression and ULBP molecule expression (30) . On the other hand, in the present study we could establish a close correlation between ULBP expression at the cell surface (as determined by the various anti-ULBP mAbs) and the susceptibility of tumor cells to NKG2D-dependent killing. Altogether, the above data indicate that the use of specific mAbs is crucial not only for the determination of MICA and ULBP expression but also to establish the surface density of these molecules.

In conclusion, the relevance of NKG2D in human NK cell triggering is strengthened by the present finding that this receptor is implied not only in killing of MICA⁺ carcinoma and melanoma cells but also of ULBP⁺ MICA⁻ tumor cells. Moreover, our results also suggest that NKG2D-induced NK cell triggering is likely to be confined to the interactions between this receptor and MICA or ULBP ligands because its involvement in NK cell-mediated cytotoxicity strictly correlates with the expression of these ligands on target cells. Along this line, when tumor target cells coexpress MICA and ULBP, both of these molecules appear to participate in NK cell activation via NKG2D. It remains to be established whether additional NKG2D ligands may exist; however, their surface expression may not be characteristic of the tumor cell lines analyzed in this study. A previous report suggested a role of NKG2D also in the NK cell-mediated recognition and killing of allogeneic phytohemagglutinin-induced blasts (22) . Therefore, it is possible that these cells may express still-undefined surface molecules that serve as additional ligands for NKG2D.